{"id":36047,"date":"2024-11-13T11:10:10","date_gmt":"2024-11-13T03:10:10","guid":{"rendered":"https:\/\/linguaresources.com\/?p=36047"},"modified":"2024-11-13T11:10:10","modified_gmt":"2024-11-13T03:10:10","slug":"cohere-releases-aya-expanse-multilingual-ai-models","status":"publish","type":"post","link":"https:\/\/linguaresources.com\/?p=36047","title":{"rendered":"Cohere Releases \u201cAya Expanse\u201d Multilingual AI Models"},"content":{"rendered":"\n<p><a href=\"https:\/\/cohere.com\/research\/aya\" target=\"_blank\" rel=\"noreferrer noopener\">Cohere for AI<\/a>, the research arm of the language&nbsp;<a href=\"https:\/\/multilingual.com\/tag\/ai\/\">artificial intelligence (AI)<\/a>&nbsp;company, has introduced two new large language models (LLMs) \u2014 Aya Expanse 8B and 32B \u2014 as part of its ongoing&nbsp;project aimed at closing language divides in foundational AI datasets and models. The Aya Expanse models provide researchers access to advanced AI capabilities across 23 languages, including Arabic, Chinese, French, and Hindi.<\/p>\n\n\n\n<p>\u201cBuilding on more than two years of open science research, Aya Expanse offers significant performance advances, setting a new state-of-the-art for multilingual LLMs,\u201d the&nbsp;<a href=\"https:\/\/cohere.com\/research\/aya\" target=\"_blank\" rel=\"noreferrer noopener\">Cohere website<\/a>&nbsp;states. \u201cThis includes a series of breakthroughs in data arbitrage, preference training for performance and safety, and model merging.\u201d<\/p>\n\n\n\n<p>According to&nbsp;<a href=\"https:\/\/siliconangle.com\/2024\/10\/24\/cohere-announces-aya-expanse-multilingual-ai-model-family-researchers\/\" target=\"_blank\" rel=\"noreferrer noopener\">a&nbsp;<em>SiliconANGLE<\/em>&nbsp;article<\/a>, the two Aya Expanse models were launched with open weights on hosting sites&nbsp;<a href=\"https:\/\/huggingface.co\/CohereForAI\/aya-expanse-8b?ref=cohere-ai.ghost.io\" target=\"_blank\" rel=\"noreferrer noopener\">Hugging Face<\/a>&nbsp;and&nbsp;<a href=\"https:\/\/www.kaggle.com\/models\/cohereforai\/aya-expanse\" target=\"_blank\" rel=\"noreferrer noopener\">Kaggle<\/a>, and they used \u201cseveral new core research innovations\u201d to achieve high performance, including \u201csynthetic data and human feedback in late-term training.\u201d<\/p>\n\n\n\n<p>In a&nbsp;<a href=\"https:\/\/cohere.com\/blog\/aya-expanse-connecting-our-world\" target=\"_blank\" rel=\"noreferrer noopener\">blog post<\/a>, Cohere claims that&nbsp;Aya Expanse 32B outperforms models like&nbsp;<a href=\"https:\/\/multilingual.com\/tag\/Google\" target=\"_blank\" rel=\"noreferrer noopener\">Google<\/a>\u2019s Gemma 2 27B and&nbsp;<a href=\"https:\/\/multilingual.com\/tag\/Meta\">Meta<\/a>\u2019s Llama 3.1 70B. For lower-parameter options, Aya Expanse 8B also demonstrated advantages over other similar-sized models like Gemma 2 9B and Llama 3.1 8B.&nbsp;\u201cThe improvements in Aya Expanse are the result of a sustained focus on expanding how AI serves languages around the world by rethinking the core building blocks of machine learning breakthroughs,\u201d the blog post states.<\/p>\n\n\n\n<p>According to&nbsp;<a href=\"https:\/\/venturebeat.com\/ai\/cohere-launches-new-ai-models-to-bridge-global-language-divide\/\" target=\"_blank\" rel=\"noreferrer noopener\">a&nbsp;<em>VentureBeat<\/em>&nbsp;article<\/a>&nbsp;by Emilia David, the Aya initiative attempts to solve the problem of research being done on LLMs that don\u2019t perform well in languages other than English.&nbsp;\u201cMany LLMs eventually become available in other languages, especially for widely spoken languages, but there is difficulty in finding data to train models with the different languages,\u201d David writes. \u201cIt can also be difficult to accurately benchmark the performance of models in different languages because of the quality of translations.\u201d<\/p>\n\n\n\n<p>Aya, derived from the Twi language term for \u201cfern,\u201d has grown into one of the world\u2019s largest open-source multilingual projects, featuring over 513 million data points curated across 101 languages and 250 language ambassadors worldwide. This collaborative approach allows Aya\u2019s datasets to expand research opportunities in regions where non-English AI resources remain limited.<\/p>\n","protected":false},"excerpt":{"rendered":"<p>Cohere for AI, the research arm of the language&nbsp;ar [&hellip;]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[391],"tags":[],"class_list":["post-36047","post","type-post","status-publish","format-standard","hentry","category-391"],"_links":{"self":[{"href":"https:\/\/linguaresources.com\/index.php?rest_route=\/wp\/v2\/posts\/36047"}],"collection":[{"href":"https:\/\/linguaresources.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/linguaresources.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/linguaresources.com\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/linguaresources.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=36047"}],"version-history":[{"count":0,"href":"https:\/\/linguaresources.com\/index.php?rest_route=\/wp\/v2\/posts\/36047\/revisions"}],"wp:attachment":[{"href":"https:\/\/linguaresources.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=36047"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/linguaresources.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=36047"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/linguaresources.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=36047"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}